The Good
OpenAI shipped a trio of new realtime voice models on 7 May: GPT-Realtime-2 (GPT-5-class reasoning in live voice), GPT-Realtime-Translate (live speech translation across 70+ input languages into 13 output languages, keeping pace with the speaker), and GPT-Realtime-Whisper (streaming transcription as you speak). A voice agent can take a real phone call, look at a document you send mid-call, and act on internal systems while you're still talking. 
Comment: Two things to register here. First, real-time multilingual voice translation at sub-dollar-per-minute economics quietly removes one of the oldest barriers in human collaboration; a Sydney call centre can now serve a Manila supplier and a Jakarta customer in their own languages without staffing it. Second, voice agents now have a phone number. The implication for any business with inbound customer service, after-hours support, or outbound calling motion is that the cost curve just went down again.

The Bad
Commonwealth Bank's ~$1bn AI-assisted loan fraud probe escalated again this fortnight, with a Melbourne accountant charged and CBA referring multiple brokers and accountants to police. The mechanism: shell companies submitted home-loan applications backed by AI-generated payslips, tax returns, and income statements good enough to clear both broker channels and CBA's own introducer scheme. ASIC, AUSTRAC and NSW Police are all involved, with aggregators including Finsure drawn in too. 
Comment: This is a taste of what industrialised AI fraud actually looks like in Australia – not deepfake calls, just convincing-PDFs-at-scale. Workflows built on “we verify the document” are challenged, the new control is verifying the source (ATO, employer, bank-feed). We’ve seen this for some time in invoicing, with payable links now the norm through Xero and other providers.

& The Ugly
US prosecutors this month unsealed charges tied to the Shunda compound in Myanmar after KNLA forces seized it and handed investigators 8,000 phones and 1,500 computers. Inside, prosecutors found a dedicated “AI room”: a female operator running live deepfake video calls on demand so male scammers could impersonate the women in stolen profile photos. Roughly 200,000 people, many trafficked, work across these compounds. Chainalysis's 2026 report puts the average AI-assisted scam at 4.5× more profitable than its pre-AI equivalent. 
Comment: A couple of ugly things at play here. Firstly, this is the new shape of the human trafficking industry and this Bloomberg exposé is worth the watch. Secondly, deepfakes have left the demo stage. It’s a scary new world, but thankfully it’s possible to guard against. I just spoke about this with my parents, and we collaborated on the questions to ask and passphrases to use if someone calls with our voices asking for money or talking about investments. 
A simple tip: if ✌️your mum✌️ calls asking for money, tell her, “No problem, I’ll call you right back and make the transfer.” If her number is being spoofed, calling back will be a reset, and the scammer will be lost in the ether.
It’s an uncomfortable conversation, but could save you or your family thousands of dollars, and a whole lot of embarrassment. Try it at your next dinner.

The “US versus China AI race” is one of the most repeated headlines going around. But what do we actually know about what's happening on the ground in China – beyond the geopolitics? I sat down with Ray Sun to get past the headlines and understand the landscape from the people and companies actually building in it.

JK: Not everyone knows the Chinese tech landscape. If readers know nothing beyond TikTok and Huawei, who should they know about?

RS: I'll do the US analogies. Amazon is to Alibaba. Meta is sort of to Tencent – they run WeChat, the main messaging platform. Google is to Baidu, the search engine. Uber is to Didi and Meituan – Didi does the cars, Meituan does the food delivery. Huawei is kind of like Apple, Ericsson and Nokia all combined. They do everything. Xiaomi is basically another Samsung – phones, cars, vacuum cleaners. It's a Dyson and a car company and a phone company combined. China's similar to South Korea with the ‘chaebol’ [rich family conglomerates’] thing. These are conglomerates, not single-product companies.

Then there's what's called the Four Dragons and the Six Tigers. The Four Dragons focused on pre-generative AI. Facial recognition, recommendation algorithms. Those companies are Alibaba, Tencent, Baidu and ByteDance. The Six Tigers are the foundation model labs: MiniMax, Moonshot (which provides Kimi), Baichuan, 01.ai, StepFun and DeepSeek. These are the ones driving post-generative AI. Similar to OpenAI, Anthropic, xAI.

The incumbent Dragons are mostly based in Shenzhen. That’s like the OG Silicon Valley of China. The new Silicon Valley is Hangzhou, close to Shanghai. Hangzhou's got really good weather and, in China, it's actually hard to find that. The local government was probably the first to recognise the importance of AI unicorns. The talent is there because the local university is kind of like the MIT of China. The regulators are considered more chill compared to other cities. And once you have one cool company there – Alibaba has their headquarters in Hangzhou – people go, workers go, startups branch out. DeepSeek is also in Hangzhou. So that's created even more buzz.

JK: I saw in the news last year that Xi Jinping gathered the big tech CEOs in one room. What happened?

RS: Yeah, so Xi gathered all the big tech CEOs. You see the seats arranged in rows, and the front row is normally reserved for Alibaba, Baidu – the Four Dragons. But the DeepSeek CEO was in the front row. This young guy, barely a few years old as a company, in his 30s or 40s, seated above other seniors. In Chinese culture, people take notice of this stuff. It's huge.

What that signals is that software – historically, China's focused on physical manufacturing, anything tangible – is now being treated as equally important. A software-driven company as important as your car manufacturer, as important as your chemicals companies. That's a massive shift in where the government's thinking is heading.

JK: In Australia, AI is largely met with distrust and pessimism. Even the latest survey results out of the Tech Policy Design Institute saw 85% of Australians support AI regulation and just 1% trust AI. What's it like on the ground in China?

RS: So for background, I do a lot of China-based AI work in my day job as a lawyer. I work with local partners in China to advise Chinese AI companies, draft contracts, and have been to China to sit  in boardrooms with legal counsel and directors. So I’m going by my  on-the-ground observations.

The software market is really tough in China and typically not considered as a ‘top’ industry, at least at a B2B level. It is really hard to sell software there. Apart from Microsoft, every other US company that tried to go into China failed. The Chinese consumer mindset is, “Software is just a bunch of files and code. You can copy and paste, it should be free.” Desktop software is a whole different game to mobile apps. It took a while for the market to even understand cloud. If you can't sell local software, how are you going to convince people to outsource to the cloud?

JK: So this explains China’s obsession with open source AI. You've told me before that in China, if it's not open source, it's a scam.

RS: For developers, open source is kind of like a ticket to your credibility. It is ‘proof of work’. There are so many options in the market that Chinese consumers naturally ask, “How do I know you're legit?” Open source libraries help, but also the reputation and profile of the founders. In China, founder integrity – the character, the personality – has a huge impact, and very much intrinsically linked to the company brand.

Any signs of gloating or pride, that gives people the ‘ick’. And the ick actually has a boycotting effect on your services. Open source is part of that. If you're open source, it means you're doing a public service “for the people” [wèi rénmín fúwù “为人民服务”]. It is a sign of transparency as consumers can see how technically sophisticated you are. You're not just someone who forked a GitHub repo. You actually created this from scratch, backed up with hardcore maths - a sign of original homegrown innovation. And people love that, especially since STEM innovation is an increasingly important pillar of national pride.

JK: Is AI just part of the culture at this point?

RS: In China, maths and science is sexy. When I was in a hotel watching TV, every fourth channel was some tech or science documentary. There's a trend where parents used to send kids to Olympiad maths classes. That was the status symbol. Then it shifted to coding classes. Now the new trend is robotics classes. Kids are going to coding and robotics classes after school, and that's as commonplace as Australian parents sending their kids to swimming. A lot of people love maths and science. Their minds are naturally geared towards AI use cases. Innovation just becomes a given.

It comes down to education. Education influences culture, culture reinforces education. That's something we could really think about here in Australia. Make science and maths cool. Children's TV shows. I think that's a way to start.

JK: I’ve heard that in Chinese society there’s a feeling like being an NPC [non-playable character] versus a main protagonist. Can you explain that?

RS: People who leave China and immigrate, that tends to be the main cultural reason. I don’t think it is controversial to say that being in a large population country and highly competitive talent market like China makes you feel like an NPC, rather than a main protagonist. It’s a hypercompetitive culture called ‘nei juan’ [involution 内卷], though South Korea, Japan and Singapore arguably have a roughly similar culture. People who want to be the protagonist, that's when they go to the West, which celebrates – and has room for – individualistic culture. But many people in China value stability. It's such a massive population that's gone through centuries of famine and poverty. To have some stability and wealth being generated, and a happy family, people see that as a blessing and I don’t blame them.

But with Gen Z, there’s that mix of westernised “main-protagonist” thinking coming through. And then there's ‘tang ping’ [“lying flat”] where some young people are just quitting the rat race and staying at home as a personal rebellion against the establishment.

JK: How much do Chinese people actually think about the AI race in terms of US versus China?

RS: Honestly, the US versus China framing tends to be overplayed in the West. When you visit China, many businesses over there don't really think about the US that much. They only care about their next-door neighbour. Each Chinese business cares more about whether the competitor next door is going to undercut you, poach your top talent or put you out of business with a new superior product…

Read the rest of the interview here.

If you’ve made it this far, thanks for joining us again. We’d love to know your thoughts – please hit us up. Otherwise, see you in the next drop.

Yours in humanity,